OrangeGrid - High-Throughput Computing

Cluster Type: High-Throughput Computing (HTC) Scheduler: HTCondor

What is OrangeGrid?

OrangeGrid is Syracuse University’s largest computing cluster with over 80,000 cores, optimized to perform a large number of parallel jobs and provide high processing capacity over extended periods. The cluster utilizes a mixture of dedicated and scavenged compute nodes managed by HTCondor.

Best For:

  • Many independent jobs (hundreds to thousands)
  • Parameter sweeps and batch processing
  • Embarrassingly parallel workloads
  • Single-node high-memory or GPU jobs
  • LLM inference across multiple inputs

Technical Overview

Scale: Over 80,000 cores across heterogeneous pool of compute nodes

CPU Architecture:

  • Mix of AMD and Intel processors
  • Varying core counts and configurations
  • Opportunistic scheduling matches jobs to available resources

GPU Capabilities:

  • GPU nodes available through SUrge infrastructure
  • Models include: NVIDIA A100, L40S, A6000, and others
  • Request with +request_gpus = 1 in submit files
  • Best for single-node GPU jobs (training, inference, rendering)

Job Characteristics:

  • Single-node jobs (one job per machine)
  • High-core count available (up to 64+ cores per node on some machines)
  • High-memory nodes available (256GB+ on select nodes)
  • Fair-share scheduling across all users

Storage: 4TB home directory (default), NetApp-based, automatically mounted on compute nodes

Access: SSH via its-og-loginX.syr.edu

How It Works

Submit jobs using HTCondor submit files. Each job runs independently on an available node in the pool:

  1. Create your script and HTCondor submit file
  2. Submit with condor_submit job.sub
  3. HTCondor matches your job to available resources
  4. Job runs when resources become available
  5. Results saved to your home directory

Typical Use Cases

  • Processing thousands of images with the same pipeline
  • Running parameter sweeps for optimization studies
  • LLM inference or fine-tuning across multiple datasets
  • Batch data analysis (R, Python, Julia)
  • Rendering farms (Blender, etc.)
  • Monte Carlo or stochastic simulations

Learn More

📊 OrangeGrid Technical Specifications
💻 OrangeGrid Code Examples

Table of contents